Generalized Electric Medicine Group Publishes Three New Research Achievements in Leading International Journals
The Generalized Electric Medicine Group is pleased to announce three new research achievements published in leading international journals, including Medical Image Analysis, Radiology: Artificial Intelligence, and IEEE Transactions on Image Processing. These publications mark another important milestone for the group in advancing intelligent medical image analysis, multimodal learning, and AI-powered clinical applications.
Together, these studies demonstrate the group’s sustained commitment to developing innovative, clinically meaningful, and technically rigorous artificial intelligence methods for next-generation healthcare.
1. Advancing Personalized Breast Cancer Risk Prediction Medical Image Analysis
Published in Medical Image Analysis, the paper “Incorporating global-local tissue changes to predict future breast cancer from longitudinal screening mammograms” presents a novel framework for predicting future breast cancer risk from longitudinal screening mammograms. Rather than relying solely on static imaging information, this work captures both global and local tissue changes over time, enabling more accurate modeling of cancer risk progression. By integrating temporal information from prior screening exams, the proposed method improves both risk stratification and time-to-event prediction, moving breast cancer screening closer to a more personalized and preventive paradigm.
This study highlights the group’s efforts to leverage longitudinal imaging data for early cancer risk assessment and to support smarter, more individualized screening strategies.
2. Building Explainable Multimodal Foundation Models for Mammography Radiology: Artificial Intelligence
In Radiology: Artificial Intelligence, researchers from the Generalized Electric Medicine Group published “Visualizing Radiologic Connections: An Explainable Coarse-to-Fine Foundation Model with Multiview Mammograms and Associated Reports.” This work introduces an explainable multimodal foundation model that jointly learns from multiview mammograms and their corresponding radiology reports. By combining imaging and textual information in a unified framework, the model acquires richer clinically relevant representations while also improving interpretability.
Importantly, the study goes beyond performance alone. It emphasizes explainable AI by visualizing the learned radiologic connections between image regions and report content, helping bridge the gap between model predictions and clinical understanding. The work demonstrates the promise of foundation-model-based approaches for improving malignancy classification, localization, and broader downstream tasks in breast imaging.
3. Enabling More Accurate MR-Only Radiotherapy with Anatomy-Aware Image Translation IEEE Transactions on Image Processing
The group’s third paper, “Anatomy-Aware MR-Imaging-Only Radiotherapy,” published in IEEE Transactions on Image Processing, addresses a critical challenge in MR-only radiotherapy planning: generating synthetic CT images from MR scans with high anatomical fidelity. While MR imaging provides excellent soft-tissue contrast, it does not directly provide the electron density information required for radiotherapy dose calculation. Existing MR-to-CT translation methods often require separate models for different anatomical regions, limiting flexibility and scalability.
To overcome this limitation, the study proposes a unified prompt-driven model that dynamically adapts to different anatomical regions and generates CT images with stronger structural consistency. The method incorporates a region-specific attention mechanism, including a region-aware vector and a dynamic gating factor, enabling a single framework to perform MRI-to-CT translation across multiple anatomical sites. Experimental and dosimetric results show that the proposed model produces clearer, more anatomically detailed CT images and achieves dose distributions that are more closely aligned with real CT-based planning. This work demonstrates the group’s strength in translating advanced image synthesis methods into clinically relevant radiotherapy solutions.
These three publications reflect the collective efforts of the Generalized Electric Medicine Group, its collaborators, and its clinical partners. Although they span different application scenarios, all three studies are united by a common goal: to create AI systems that are not only methodologically advanced, but also trustworthy, interpretable, and clinically impactful.
From longitudinal cancer risk prediction, to explainable multimodal foundation modeling, to anatomy-aware radiotherapy imaging, these achievements showcase the breadth and depth of the group’s research at the intersection of artificial intelligence and medicine.
The Generalized Electric Medicine Group will continue to push the frontiers of medical AI by developing innovative computational methods that support early diagnosis, precision treatment, and improved patient outcomes.